[LU-3873] HSM cancel actions never removed from agent llog

• Status: Open
• Project: Lustre
• Affects Version/s: Lustre 2.5.0
• Type: Bug
• Priority: Major
• Reporter: John Hammond
• Assignee: WC Triage
• Resolution: Unresolved
• Labels: HSM
• Severity: 3

Description

Cancel actions are ignored by the copytool and hence never get beyond the ARS_STARTED status. Despite stopping everything, remounting, and restarting I still show cancel actions:

q:~# cat /proc/fs/lustre/mdt/lustre-MDT0000/hsm/agent_actions
lrh=[type=10680000 len=136 idx=8] fid=[0x200000400:0x2:0x0] dfid=[0x200000400:0x2:0x0] compound/cookie=0x5226288f/0x522610fc action=CANCEL archive#=0 flags=0x0 extent=0x0-0xffffffffffffffff gid=0x0 datalen=0 status=STARTED data=[]

Shouldn't the CT respond to every action?

Comments

[jacques-charles lafoucriere]

CT support for CANCEL is optionnal. The CANCEL request is set SUCCEED when the CT after sending a progress has been notifyed by CDT. It is removed from llog based on expiration time. After a remount we can clean them to avoid waiting for the expiration time

[Bruno Faccini (Inactive)]

Sorry, I am VERY late on this ticket.
And re-reading it, I wonder if it is a real problem or not? Is there any issue implied if these actions wait for the expiration time to clean-up ?

[Aurelien Degremont (Inactive)]

This is more an optimization rather than a real bug. It could be a little bit confusing to admins thought.
It seems the request llog could be a little bit too small sometimes and removing this useless records could may be help.

I think we need to improve this with a better management of this, but this ticket has a low priority IMHO.